Welcome to the ExecuTorch Documentation

Important

This is an alpha release; the ExecuTorch APIs and the .pte binary format may change in incompatible ways before stabilizing in a future beta release. When deploying models, we currently recommend using a version of the runtime built from the same git revision that was used to generate the .pte file. Once the format has stabilized, this will no longer be necessary.

We welcome any feedback, suggestions, and bug reports from the community to help us improve the technology. Please use the PyTorch Forums for discussion and feedback about ExecuTorch using the ExecuTorch category, and our GitHub repository for bug reporting.

ExecuTorch is a PyTorch platform that provides infrastructure to run PyTorch programs everywhere from AR/VR wearables to standard on-device iOS and Android mobile deployments. One of the main goals for ExecuTorch is to enable wider customization and deployment capabilities of the PyTorch programs.

ExecuTorch heavily relies on such PyTorch technologies as torch.compile and torch.export. If you are not familiar with these APIs, you might want to read about them in the PyTorch documentation before diving into the ExecuTorch documentation.

The ExecuTorch source is hosted on GitHub at https://github.com/pytorch/executorch.

Getting Started

Topics in this section will help you get started with ExecuTorch.

What is ExecuTorch?

A gentle introduction to ExecuTorch. In this section, you will learn about main features of ExecuTorch and how you can use them in your projects.

Getting started with ExecuTorch

A step-by-step tutorial on how to get started with ExecuTorch.

ExecuTorch Intermediate Representation API

Learn about EXIR, a graph-based intermediate representation (IR) of PyTorch programs.

Tutorials and Examples

Ready to experiment? Check out some of the ExecuTorch tutorials.

---

Exporting to ExecuTorch Tutorial

Topics: Export,Delegation,Quantization

A tutorial for exporting a model and lowering a it to be runnable on the ExecuTorch runtime.

Running an ExecuTorch Model C++ Tutorial

Topics:

A tutorial for setting up memory pools, loading a model, setting inputs, executing the model, and retrieving outputs on device.

Simplified Runtime APIs Tutorial

Topics:

A simplified tutorial for executing the model on device.

Using the ExecuTorch SDK to Profile a Model

Topics: SDK

A tutorial for using the ExecuTorch SDK to profile and analyze a model with linkage back to source code.

Building an ExecuTorch iOS Demo App

Topics: Delegation,iOS

A demo tutorial that explains how to build ExecuTorch into iOS frameworks and run an iOS app.

Building an ExecuTorch Android Demo App

Topics: Delegation,Android

A demo tutorial that explains how to build ExecuTorch into a JNI library and build an Android app.

Lowering a Model as a Delegate

Topics: Export,Delegation

An end-to-end example showing how to lower a model as a delegate

Building and Running ExecuTorch with XNNPACK Backend

Topics: Export,Backend,Delegation,Quantization,XNNPACK

A demo tutorial for lowering and exporting models with the XNNPACK Backend

Building and Running ExecuTorch with Vulkan Backend

Topics: Export,Backend,Delegation,Vulkan

A tutorial that walks you through the process of building ExecuTorch with Vulkan Backend

Building and Running ExecuTorch with ARM Ethos-U Backend

Topics: Export,Backend,Delegation,ARM,Ethos-U

A tutorial that walks you through the process of building ExecuTorch with ARM Ethos-U Backend

Building and Running ExecuTorch with CoreML Backend

Topics: Export,Backend,Delegation,CoreML

A tutorial that walks you through the process of building ExecuTorch with CoreML Backend

Building and Running ExecuTorch with MPS Backend

Topics: Export,Backend,Delegation,MPS,MPSGraph

A tutorial that walks you through the process of building ExecuTorch with MPSGraph Backend

Building and Running ExecuTorch with Qualcomm AI Engine Direct Backend

Topics: Export,Backend,Delegation,QNN

A tutorial that walks you through the process of building ExecuTorch with Qualcomm AI Engine Direct Backend

Building and Running ExecuTorch on Xtensa HiFi4 DSP

Topics: Export,Custom-Operators,DSP,Xtensa

A tutorial that walks you through the process of building ExecuTorch for an Xtensa Hifi4 DSP using custom operators